Metraipolis › archive › dc_16f019d83c03

Faster Block-Diffusion Serving with Distribution-Free Risk Guarantees — arXiv:2609.33887

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.33887
License (as archived)
CC0-1.0
Archived
2026-10-07T21:53:12.733Z
SHA-256 of the archived text
1f30506c23eb28ccd49308c09a758d50f51d15ad9412ef3f30b53de130d4b79a
Stable URL of this page
https://metraipolis.com/documento/dc_16f019d83c03

How to cite

Faster Block-Diffusion Serving with Distribution-Free Risk Guarantees — arXiv:2609.33887 — arXiv (en), https://arxiv.org/abs/2609.33887 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_16f019d83c03 (Metraipolis archive dc_16f019d83c03, sha256 1f30506c23eb28ccd49308c09a758d50f51d15ad9412ef3f30b53de130d4b79a)
BibTeX
@misc{dc16f019d83c03,
  title = {Faster Block-Diffusion Serving with Distribution-Free Risk Guarantees — arXiv:2609.33887},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.33887},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_16f019d83c03; sha256 1f30506c23eb28ccd49308c09a758d50f51d15ad9412ef3f30b53de130d4b79a},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33887. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Block-diffusion language models are served at hand-picked operating points, such as acceptance thresholds, buffer depth, schedule, checkpoint and precision, and each point is chosen by its mean benchmark accuracy. However, a mean does not tell an operator how often a faster configuration fails on prompts that the slower one answers correctly. On the serving engine and its decode traces, the default commit rule already commits every fully resolved block, a static skip rule captures nearly all of the compute that allocation can save, and self-distillation on engine-decoded targets adds speed at unchanged accuracy. Larger speedups come from lower thresholds, which commit tokens that are still uncertain. We therefore present Redline, a finite-sample procedure that selects operating points, hand-picked or learned, from the correctness of their answers on calibration prompts. Redline keeps the reference-relative risk, the joint probability that the reference answers correctly and a candidate configuration does not, within a user-chosen budget with high probability, and deploys the fastest configuration that passes. It speeds up math at a smaller risk budget than code in both model families, and at a budget of ten percent it deploys a LLaDA2 math configuration that commits over a third more tokens in each forward. It also applies without modification to the acceptance rule of speculative decoding and to weight quantization. On the same calibration data, Redline stays within its stated failure probability, whereas each tolerance of a mean-accuracy rule either gains less speed for some model and task or exceeds the risk budget far more often for another. Code is available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.33887 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T21:27:08.564Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).